| 摘要: |
| 针对在复杂电磁环境中物理层无线入侵检测受时变信道的影响导致精度不高的问题，提出了一种基于电磁地图的物理层无线入侵检测策略。该策略首先利用径向基函数（Radial Basis Function，RBF）神经网络优化后的随机森林算法对残缺的电磁数据进行插值处理，构建电磁地图；然后提取电磁地图中所蕴含的辐射源的信道状态信息（Channel State Information，CSI）的相位差并将其转换为热力图，并利用WGAN-Encoder模型对热力图进行重构和映射，结合输出的特征残差和热力图重建误差得到的异常分数实现入侵辐射源身份识别；最后通过实时更新电磁地图实现辐射源检测与跟踪，进而完成物理层入侵检测。实验结果表明，该策略的曲线下面积(Area Under Curve，AUC)和等错误率(Equal Error Rate，EER)分别为0.988和0.084。 |
| 关键词: 复杂电磁环境 物理层无线入侵检测 电磁地图 RBF神经网络 随机森林 |
| DOI：10.20079/j.issn.1001-893x.250513002 |
|
| 基金项目:国家自然科学基金面上项目（61971473）；湖南省大学生创新训练项目（S202490002323） |
|
| A Strategy of Physical-layer Wireless Intrusion Detection Based on Electromagnetic Maps |
| ZHANG Jiaqi,WANG Yuanjie,HAN Ruichang,WANG Hongjun |
| (College of Electronic Countermeasure，National University of Defense Technology，Hefei 230031，China) |
| Abstract: |
| In response to the problem of low accuracy caused by the time-varying channels in physical-layer wireless intrusion detection in complex electromagnetic environment，a strategy of physical-layer wireless intrusion detection based on electromagnetic maps is proposed，in which a random forest algorithm optimized by radial basis function(RBF) neural network is adopted to interpolate incomplete electromagnetic data and construct electromagnetic maps.Then the channel state information(CSI) phase difference of radiation sources in the electromagnetic map construction is extracted and converted into the heatmap.WGAN-Encoder model is introduced to reconstruct and map the heatmap and combine with the anomaly score calculated from feature residuals and heatmap reconstruction errors to realize authentication of intrusion radiation sources.Finally，the real-time update electromagnetic map is used to detect and track the location of radiation sources，achieving physical-layer wireless intrusion detection.The results of area under curve(AUC) and equal error rate(EER) of the intrusion detection strategy are 0.988 and 0.084，respectively. |
| Key words: complex electromagnetic environment physical-layer wireless intrusion detection electromagnetic map RBF neural network random forest |
《电讯技术》编辑部 